Molecular Phylogenetics and Evolution 113 (2017) 150–160

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

journal homepage: www.elsevier .com/locate /ympev

Reconstructing the molecular phylogeny of giant sengis (Macroscelidea;Macroscelididae; Rhynchocyon)

http://dx.doi.org/10.1016/j.ympev.2017.05.0121055-7903/� 2017 Elsevier Inc. All rights reserved.

Abbreviations: 12s16s, 12S rRNA, valine tRNA, and 16S rRNA; AMNH, AmericanMuseum of Natural History; BLAST, basic local alignment search tool; BMNH,Natural History Museum, London; Bpp, Bayesian posterior probability; CAS,California Academy of Sciences; CASMAM, California Academy of Sciences Mam-malogy; D-loop, hypervariable 50 end of the control region; ENAM, Enamelin;FMNH, Field Museum of Natural History; IRBP, inter-photoreceptor retinoid-binding protein; MCMC, Markov Chain Monte Carlo; MCZ, Museum of ComparativeZoology; mlb, maximum likelihood bootstrap; MTSN, Museo Tridentino di ScienzeNaturali; NCBI, National Center for Biotechnology Information; ND2, NADHdehydrogenase 2; PCR, polymerase chain reaction; RAxML, Random AxeleratedMaximum Likelihood; RMCA, Royal Museum of Central Africa; SNP, singlenucleotide polymorphism; UAM, University of Alaska Museum; vWF, Von Wille-brand factor.⇑ Corresponding author.

E-mail address: jdumbacher@calacademy.org (J.P. Dumbacher).1 Present address: Department of Biology, Fordham University, Bronx, NY 10458,

United States.

Elizabeth J. Carlen a,b,1, Galen B. Rathbun a, Link E. Olson c, Christopher A. Sabuni d, William T. Stanley e,John P. Dumbacher a,b,⇑a Institute for Biodiversity Science and Sustainability, California Academy of Sciences, San Francisco, CA 94118, United StatesbDepartment of Biology, San Francisco State University, San Francisco, CA 94132, United StatescUniversity of Alaska Museum, Fairbanks, AK 99775, United Statesd Pest Management Centre, African Centre of Excellence, Sokoine University of Agriculture, Morogoro, Tanzaniae Science & Education, Field Museum of Natural History, Chicago, IL 60605, United States

a r t i c l e i n f o

Article history:Received 14 February 2017Revised 5 May 2017Accepted 12 May 2017Available online 24 May 2017

Keywords:RhynchocyonGiant sengisElephant-shrewsAfricaMacroscelididaePhylogeneticsTaxonomy

a b s t r a c t

Giant sengis (Macroscelidea; Macroscelididae; Rhynchocyon), also known as giant elephant-shrews, aresmall-bodied mammals that range from central through eastern Africa. Previous research on giant sengisystematics has relied primarily on pelage color and geographic distribution. Because some species havecomplex phenotypic variation and large geographic ranges, we used molecular markers to evaluate thephylogeny and taxonomy of the genus, which currently includes four species: R. chrysopygus, R. cirnei(six subspecies), R. petersi (two subspecies), and R. udzungwensis. We extracted DNA from fresh and his-torical museum samples from all taxa except one R. cirnei subspecies, and we generated and analyzedapproximately 4700 aligned nucleotides (2685 bases of mitochondrial DNA and 2019 bases of nuclearDNA) to reconstruct a molecular phylogeny. We genetically evaluate Rhynchocyon spp. sequences previ-ously published on GenBank, propose that the captive R. petersi population in North American zoos islikely R. p. adersi, and suggest that hybridization among taxa is not widespread in Rhynchocyon. TheDNA sample we have from the distinctive but undescribed giant sengi from the Boni forest of northerncoastal Kenya is unexpectedly nearly identical to R. chrysopygus, which will require further study. Ouranalyses support the current morphology-based taxonomy, with each recognized species forming amonophyletic clade, but we propose elevating R. c. stuhlmanni to a full species.

� 2017 Elsevier Inc. All rights reserved.

1. Introduction restricted to the African continent and form two well-defined sub-

The 19 extant species of sengis (elephant-shrews; Rathbun andKingdon, 2006) in the mammalian Order Macroscelidea are

families, the soft-furred sengis (Macroscelidinae), with 15 extantspecies in four genera (Elephantulus, Macroscelides, Petrodromus,and Petrosaltator), and the giant sengis (Rhynchocyoninae), withfour extant species in one genus (Rhynchocyon).

Despite their long evolutionary history (Novacek, 1984) andbroad African distribution (Corbet and Hanks, 1968) in highlydiverse habitats across much of Africa (Rathbun, 2009), sengishave proven to be taxonomically challenging, having relativelyfew discretely varying morphological traits with which to resolvetheir phylogeny and taxonomy (Corbet and Hanks, 1968). Withthe application of molecular genetics in the last several decades,some insights into extant sengi phylogeny and taxonomy havebeen gained. This work has shown that Macroscelididae are mor-phologically specialized, yet across a diversity of habitats, theymaintain a stable life history and morphology that has maskedsome of their evolutionary and ecological diversity (Rathbun,2009).

E.J. Carlen et al. /Molecular Phylogenetics and Evolution 113 (2017) 150–160 151

Giant sengis, as their name indicates, are the largest members ofthe order, with body masses ranging from 300 g to 700 g. They arediurnal, swift quadrupedal forest-floor dwellers with proportion-ally long legs, a long sparsely-haired tail, and a long snout thatcan twist and probe in leaf litter in search of invertebrate prey.The golden-rumped sengi (R. chrysopygus) is the only giant sengiwhose behavioral ecology has been studied in sufficient detail toreveal that its life history is unusual for a small mammal(Rathbun, 2009). Individuals form monogamous pairs on territo-ries, shelter singly in leaf nests on the forest floor, and produceone relatively precocial offspring at a time (FitzGibbon, 1997;Rathbun, 1979).

In the 65 years between 1847 and 1912, ten species and foursubspecies of Rhynchocyon were described. Corbet and Hanks(1968), using mostly distinctive pelage color patterns (Fig. 1) and

Fig. 1. Study skins showing the color patterns of Rhynchocyon taxa (see text for museum achrysopygus CAS MAM 24526; (B) R. cirnei stuhlmanni AMNH 49462 (light form of cline)form of cline), (E) R. cirnei shirensis AMNH 161777; (F) R. cirnei cirnei CAS MAM 29358; (GRhynchocyon is not represented, but is superficially similar to R. udzungwensis, as are thestuhlmanni from western Uganda (see Corbet and Hanks, 1968).

allopatric distributions (Fig. 2), conducted a thorough taxonomicrevision of the order, resulting in only three recognized giant sengispecies. The golden-rumped sengi (R. chrysopygus) is monotypicand occurs in coastal Kenya. The black-and-rufous sengi (R. petersi),has two subspecies: R. p. adersi from islands off Tanzania and R. p.petersi from mainland Tanzania and Kenya (Fig. 2). The checkeredsengi (R. cirnei) has six subspecies: R. c. cirnei from Mozambiqueand southern Malawi, R. c. shirensis from the Shire Valley of south-ern Malawi, R. c. reichardi from Tanzania, Malawi, and Zambiahighlands, R. c. hendersoni from highlands of northern Malawi, R.c. macrurus from southeastern Tanzania lowlands, and R. c. stuhl-manni from the Congo Basin and western Uganda (Fig. 2). Rhyn-chocyon c. shirensis was a new taxon (Corbet and Hanks, 1968),whereas the other subspecies had previously been described as fullspecies. Corbet and Hanks (1968) also noted that R. c. stuhlmanni

bbreviations associated with following catalog numbers). From top to bottom: (A) R., (C) R. cirnei reichardi CAS MAM 28535; (D) R. cirnei macrurus AMNH 179301 (light) R. udzungwensis CAS MAM 28043; (H) R. petersi petersi CAS MAM 30667. The Boniclinal dark forms of R. c. macrurus from southeastern coastal Tanzania and R. cirnei

LegendLakes

Rivers

country

R. udzungwensis

R. petersi

R. cirnei

R. chrysopygus

0 125 250 375 50062.5Kilometers

Kenya

Tanzania

MozambiqueZambia

Dem Repof Congo

Ethiopia

Uganda

Angola

Lake Victoria

Sudan

Kenya

Tanzania

MAL

#

Fig. 2. Distribution of the four species of Rhynchocyon, with the polygon for R. cirnei in the Democratic Republic of Congo (Congo Basin) and western Uganda representing R.cirnei stuhlmanni, which we propose as a fifth species (see text for general distributions of other subspecies). The # is the location of the undescribed form of Rhynchocyon innorthern coastal Kenya. Distribution polygons are courtesy of IUCN (www.iucnredlist.org), which are based on point data compilation by GBR (www.sengis.org/distribution/).Collection localities (see Table 1) for each taxon are denoted by a d.

152 E.J. Carlen et al. /Molecular Phylogenetics and Evolution 113 (2017) 150–160

could arguably be elevated to full species based on its short nasalbones, all-white tail, and allopatric distribution in the Congo Basin,though they left this unresolved.

In 2008, Rovero et al. described a fourth species, R. udzungwen-sis, which occurs in two evergreen forests in the Udzungwa Moun-tains, Tanzania (Figs. 1C and 2). Andanje et al. (2010) reported apotentially new giant sengi from the Boni and Dodori nationalreserves on the northern coast of Kenya (Fig. 2) that most closelyresembles R. udzungwensis in coloration, but the phylogeny andtaxonomy remains to be studied.

Despite the seminal work by Corbet and Hanks (1968) andrecent discoveries, Rhynchocyon taxonomy remains problematic.For example, within R. cirnei, where nearly all subspecies have dis-tinct checkering patterns on the back (Fig. 1B-F), determining tax-onomic status and relationships have been difficult. Similarly,

relationships and placement of some Rhynchocyon taxa has beendifficult, especially those taxa whose checkering is masked withdark pelage (e.g., R. petersi, R. udzungwensis, Fig. 1G,H, and the darkcoastal form of R. cirnei macrurus).

A complete and accurate phylogeny for Rhynchocyon is neededfor several reasons. First, Rhynchocyon has important conservationvalue. Because Rhynchocyoninae taxa are few and distantlyrelated to other mammals (divergence approximately 42.7 ± 4.8MYA, Douady et al., 2003), they contribute significant ecologicaland phylogenetic diversity to their communities (Davies andBuckley, 2011; Faith, 1992). Many of the taxa have narrow ranges(e.g. R. udzungwensis, R. chrysopygus, R. c. shirensis), and some thatappear to have larger ranges are actually restricted to small frag-mented patches of montane and coastal forest (e.g. R. petersi).Thus, understanding how many distinct Rhynchocyon taxa exist

E.J. Carlen et al. /Molecular Phylogenetics and Evolution 113 (2017) 150–160 153

and their relationships to each other within the genus will helpdetermine how to manage populations when they becomesmaller or as some populations are inevitably extirpated. Second,there is a significant captive population of R. petersi living in zoos.These zoos maintain a coordinated breeding program withdetailed studbooks, but there is no detailed locality informationfor the founders collected from eastern Africa. For these to havemaximum conservation value for breeding and potential reintro-duction, it is imperative to know their source subspecies and pop-ulation. Finally, there are several sequences of Rhynchocyon inGenBank, and we have reason to question the accuracy of theseDNA sequence data and metadata. In particular, sequences origi-nally published as Rhynchocyon sp. (Douady et al., 2003), havebeen determined by Smit et al. (2011) to be R. chrysopygus. Basedupon the likely collecting locality, we doubted this identification,and hoped to test it both genetically and with information fromthe voucher specimen. Based upon published ribosomal sequencedata, we also doubted the accuracy of other sequences, and wehope to provide more accurate data for research. Although manysequences on GenBank have errors, these Rhynchocyon sequencesare particularly problematic because multiple studies of evolu-tionary processes have used these to represent the family or sub-family (e.g. Puttick and Thomas, 2015; Smith et al., 2013, 2016),and errors could affect both inferred relationships and evolution-ary timing and divergences.

Although several studies on the molecular phylogenetics ofsoft-furred sengis have been published (Douady et al., 2003;Dumbacher et al., 2012, 2014, 2016; Smit et al., 2007, 2008,2011), comparatively few have been done on the giant sengis(Lawson et al., 2013; Sabuni et al., 2016; Smit, 2008). Douadyet al. (2003) included a single Rhynchocyon specimen in their anal-ysis of the role of the Sahara in the diversification of Macroscelidea,but neither a voucher nor the species were identified, although thecollection locality was identified as southeastern Tanzania(Douady, 2001). Smit et al. (2011), in their study of the phyloge-netic relationships of Macroscelididae, sequenced approximately2000 bases of the mtDNA gene fragments 12S rRNA, valine tRNA,and 16S rRNA (12s16s) from one each of a R. chrysopygus, R. c.reichardi, and R. p. petersi from the Natural History Museum in Lon-don. Based on their phylogenetic analysis, Smit et al. (2011) pro-posed that R. petersi and R. cirnei were sister species, and R.chrysopygus was sister to them, and further identified the Douadyet al. (2003) Rhynchocyon sequence as R. chrysopygus. However,based on the collection locality of the Douady (2001) tissue insoutheastern Tanzania, Smit et al.’s (2011) identification seemsunlikely. Most recently, Lawson et al. (2013), examined the inter-specific relationship of R. udzungwensis and R. c. reichardi from fourforest sites, including the contact zone between the two taxa, inTanzania (Fig. 1). They analyzed three mitochondrial loci (ND2,D-loop, 12s) and two nuclear loci (ENAM, vWF) and found the indi-vidual nuclear gene trees strongly supported the monophyly of R.udzungwensis. However, due to the mixing of mitochondrial cladesamong species in their phylogeny, Lawson et al. (2013) concludedthat ancient (but not current) hybridization occurred between thetwo taxa because of the reciprocal monophyly of the nuclear allelesand because they did not find morphologically intermediatehybrids. However, it is unclear if historical introgression is wide-spread among Rhynchocyon taxa.

The objective of our research was to generate and analyze DNAsequence data for all named Rhynchocyon taxa to reconstruct phy-logenetic relationships within the genus. The phylogeny will allowus to determine the authenticity of GenBank Rhynchocyonsequences, determine the subspecies of the captive R. petersi pop-ulation, look for evidence of widespread hybridization amongRhynchocyon taxa, and assess the currently accepted taxonomy ofextant Rhynchocyon.

2. Materials and methods

2.1. Taxon sampling

We were able to obtain samples from all currently recognizedRhynchocyon taxa (Corbet and Hanks, 1968) except for R. c. hender-soni (Table 1). Fresh tissue preserved in alcohol was available withvoucher specimens in the mammalogy collections at the CaliforniaAcademy of Sciences (CAS) and the Field Museum of Natural His-tory (FMNH). Unvouchered fresh tissue was also collected for thisproject (Table 1, and Supplemental Material Table 1). We sampleddried tissue from museum study skins when fresh tissue was notavailable. Table 1 is a complete list of specimens sampled, includ-ing locality coordinates, and Fig. 2 shows the dispersion of collec-tion localities. Additionally, we used GenBank sequences fromDouady et al. (2003) and Smit et al. (2011).

2.2. DNA sequencing

We chose to study three independently segregating loci basedon previous work done with the family Macroscelididae (Douady,2001; Douady et al., 2003; Dumbacher et al., 2014; Lawson et al.,2013; Smit et al., 2011; Springer et al., 1997). We sequenced2685 bases from a mitochondrial region that includes genes for12s ribosomal RNA, tRNA-valine, and 16s ribosomal RNA (hereafter12s16s), 976 bases of the nuclear locus inter-photoreceptorretinoid-binding protein exon 1 (IRBP), and 1043 bases of thenuclear locus von Willebrand factor exon 28 (vWF).

We extracted DNA from approximately 25 mg of tissue storedin ethanol and frozen at �80 �C using a DNeasy Blood and Tissueextraction kit (Qiagen, Valencia, CA, USA). Polymerase chain reac-tion (PCR) was performed on DNA extracts using multiple primerpairs (see Supplementary Material, Table 2). For DNA extractionsfrom fresh tissue, we performed PCR in 25 ll reactions with Invit-rogen Taq (Life Technologies, South San Francisco, California, USA)and typical PCR reagents and protocols optimized for each sampletype (see Supplementary Material, Detailed Lab Methods).

For historical museum specimens, we sampled approximately25 mg of dried tissue from the hind foot or the ventral incision ofthe dried specimen. DNA was extracted in a dedicated ancientDNA laboratory at the California Academy of Sciences (CAS) orthe University of Alaska Museum (UAM). For a subset of historicalsamples, DNA extraction, PCR, and sequencing were performed inboth laboratories, providing an independent replication for thoseindividuals. Detailed protocols for historical DNA extraction andPCR can be found in Supplementary Materials, Detailed LabMethods.

PCR amplicons were Sanger sequenced using BigDye Termina-tor version 3.1 cycle sequencing chemistry (Life Technologies,South San Francisco, California, USA). Sequences were visualizedon an ABI 3130 Genetic Analyzer (Life Technologies, South SanFrancisco, California, USA) located at CAS’s Center for ComparativeGenomics or sequenced at the High Throughput Genomics Center,Seattle WA (http://www.htseq.org/).

2.3. Alignment and analysis

Because of a higher likelihood of contamination, all ampliconsfrom historical DNA were checked for contamination using theblastn, megablast, and discontiguous megablast algorithms forthe nucleotide Basic Local Alignment Search Tool (BLAST) on theNational Center for Biotechnology Information (NCBI) website.Sequences were assembled, edited, and a consensus sequence foreach individual was created in Geneious v7.1.4 (Kearse et al.,2012). For heterozygous sequences at nuclear loci, both alleles

Table 1Data for specimens used for DNA sequencing. Museum numbers are given for vouchered specimens, and field numbers are given for unvouchered specimens and denoted with anasterisk (*). Museum codes are in the abbreviations section. §Denotes sequence was downloaded from GenBank. yDenotes sequence is from historical DNA. Collection localityincludes the latitude and longitude in decimal degrees.

Taxon Voucher/field number GenBank Accession Number Collection locality

12s16s IRBP vWF

E. edwardii Unknown AY310885§ AY310899§ AY310892§ South AfricaM. micus CASMAM27997 KF895104§ KF742665§ KF742645§ Khorixas District, Kunene Region, Namibia; �20.7266, 14.1283P. tetradactylus Unknown AY310883§ AY310897§ AY310890§ Chingulungulu, TanzaniaR. chrysopygus CASMAM24525 KT348460y KT348366y &

KT358508yKT358505y &KT358506y

Gedi National Monument, Kilifi District, Kenya; �3.3097, 40.0182

R. chrysopygus CASMAM24526 KT348461y None None Gedi National Monument, Kilifi District, Kenya; �3.3097, 40.0182R. chrysopygus FMNH153106 KT348462y None None Mombasa, Kilifi District, Kenya; �4.05, 39.6667R. c. cirnei CASMAM29344 KT348463 KT348372 KT348411 Mareja Reserve, Mozambique; �12.8436, 40.1617R. c. cirnei CASMAM29345 KT348466 KT348405 KT348412 Mareja Reserve, Mozambique; �12.8483, 40.1649R. c. cirnei CASMAM29351 KT348470 KT348406 &

KT348407KT348413 Mareja Reserve, Mozambique; �12.8440, 40.1609

R. c. cirnei CASMAM29352 KT348468 KT348375 KT348414 Mareja Reserve, Mozambique; �12.8420, 40.1637R. c. cirnei CASMAM29353 KT348469 KT348376 KT348415 Mareja Reserve, Mozambique; �12.8452, 40.1615R. c. cirnei CASMAM29355 KT348464 KT348377 KT348423 Mareja Reserve, Mozambique; �12.8420, 40.1637R. c. cirnei CASMAM29357 KT348465 KT348378 KT348416 Mareja Reserve, Mozambique; �12.8429, 40.1623R. c. cirnei CASMAM29358 KT348467 KT348379 KT348417 Mareja Reserve, Mozambique; �12.8450, 40.1614R. c. macrurus RMCA 96.037-M-5388 or

RMCA 96.037-M-5390AY310880§ AY310894§ AY310887§ Chingulungulu region, Tanzania; �10.44, 38.33

R. c. macrurus FMNH88204 KT348471y None None Mihuru, Newala District, Mtwara Region, Tanzania; �10.6667, 39.5R. c. reichardi FMNH171474 KT348474 KT348400 KT348452 Mbizi Mts, Mbizi Forest Reserve, vicinity of Mazumba Hill,

Sumbawanga District, Rukwa Region, TanzaniaR. c. reichardi FMNH171617 KT348475 KT348380 KT348448 &

KT348451Mbizi Mts, vicinity of Mazumba, Sumbawanga District, Rukwa Region,Tanzania

R. c. reichardi FMNH177823 KT348476 None KT348449 Mahale Mts, Mahale National Park, 0.5 km NW Nkungwe Hill summit,Kigoma District, Kigoma Region, Tanzania; �6.1043, 29.7790

R. c. reichardi FMNH178010 KT348477 KT348381 KT348450 Mahale Mts, Mahale National Park, 0.5 km NW Nkungwe Hill summit,Kigoma District, Kigoma Region, Tanzania; �6.1043, 29.7790

R. c. reichardi(labeled R. c.hendersoni)

MCZ43732 KT348473y KT348404y KT348447y Vipya Plateau, Malawi

R. c. shirensis AMNH161777 KT348472y None None Mlanje Plateau, MalawiR. c. stuhlmanni M300* KT348478 KT348409 KT348453 Democratic Republic of the Congo; 0.0131, 25.5565R. c. stuhlmanni MK001* None KT348409 KT348454 Democratic Republic of the Congo; 0.2946, 25.2917R. petersi spp. CASMAM28767 KT348479 KT348382 KT348424 Houston Zoo, Houston, Texas, United States of AmericaR. petersi spp. CASMAM29516 KT348480 KT348383 KT348425 Houston Zoo, Houston, Texas, United States of AmericaR. p. adersi MCZ22829 KT348481y None None Nyanga Id., Zanzibar, TanzaniaR. p. petersi FMNH151213 KT348482 KT348384 KT348418 South Pare Mts, Chome Forest Reserve, 5.5 km S Bombo, near Kanza

Village, Kilimanjaro Region, Tanzania; �4.32, 38R. p. petersi FMNH151214 KT348483 KT348401 KT348419 South Pare Mts, Chome Forest Reserve, 7 km S Bombo, Kilimanjaro

Region, Tanzania; �4.33, 38R. p. petersi FMNH161311 KT348485 KT348385 KT348420 Nguru Mts, Manyangu Forest Reserve, near Disango, Morogoro

District, Morogoro Region, Tanzania; �6.04, 37.5467R. p. petersi FMNH161312 KT348486 KT348386 KT348427 Nguru Mts, Manyangu Forest Reserve, near Disango, Morogoro

District, Morogoro Region, Tanzania; �6.04, 37.5467R. p. petersi FMNH192684 KT348484 KT348402 KT348422 North Pare Mts, Minja Forest Reserve, Mwanga District, Kilimanjaro

Region, Tanzania; �3.5815, 37.6773R. p. petersi FNMH161395 KT348501 KT348373 &

KT358507KT348421 Nguru Mts, Manyangu Forest Reserve, near Disango, Morogoro

District, Morogoro Region, Tanzania; �6.04, 37.5467R. p. petersi RP15* None KT348387 KT348428 Zaraninge Forest, Tanzania; �6.1367, 38.6055R. p. petersi TA1812* None KT348374 None Zaraninge Forest, Tanzania; �6.1055, 38.6158R. p. petersi TA1818* KT348494 KT348408 KT348436 Askari Forest, Tanzania; �5.9955, 38.7607R. p. petersi TA1833* KT348487 KT348388 KT348429 Zaraninge Forest, Tanzania; �6.1056, 38.6167R. p. petersi TA1835* KT348495 KT348389 KT348430 Zaraninge Forest, Tanzania; �6.1126, 38.6211R. p. petersi TZ22766* KT348488 KT348390 KT348431 Gendagenda Forest, Tanzania; �5.5759, 38.6423R. p. petersi TZ22767* KT348499 KT348391 KT348437 &

KT348442Gendagenda Forest, Tanzania; �5.5639, 38.6502

R. p. petersi TZ22769* KT348498 KT348392 KT348426 &KT348443

Gendagenda Forest, Tanzania; �5.5871, 38.6395

R. p. petersi TZ22770* KT348500 KT348393 KT348432 Gendagenda Forest, Tanzania; �5.5871, 38.6404R. p. petersi TZ22774* KT348491 KT348394 KT348433 Kwamsisi Forest, Tanzania; �5.8909, 38.5928R. p. petersi TZ22775* KT348489 KT348403 KT348434 Kwamsisi Forest, Tanzania; �5.8921, 38.5938R. p. petersi TZ22776* KT348492 KT348395 KT348438 &

KT348444Kwamsisi Forest, Tanzania; �5.8921, 38.5939

R. p. petersi TZ22778* KT348493 KT348396 KT348439 Kwamsisi Forest, Tanzania; �5.8938, 38.5949R. p. petersi TZ22779* KT348496 KT348397 KT348440 &

KT348445Kwamsisi Forest, Tanzania; �5.8937, 38.5944

R. p. petersi TZ22783* KT348490 KT348398 KT348435 Gendagenda Forest, Tanzania; �5.601, 38.6468R. p. petersi TZ22811* KT348497 KT348399 KT348441 &

KT348446Kwamsisi Forest, Tanzania; �5.8723, 38.5726

R. udzungwensis CASMAM28043 KT348503 KT348368 KT348455 Udzungwa Mountains, Ndundulu Forest, Tanzania; �7.8045, 36.5059R. udzungwensis CASMAM28318 KT348504 KT348369 KT348456 Udzungwa Mountains, Ndundulu Forest, Tanzania; �7.7944, 36.4919

154 E.J. Carlen et al. /Molecular Phylogenetics and Evolution 113 (2017) 150–160

Table 1 (continued)

Taxon Voucher/field number GenBank Accession Number Collection locality

12s16s IRBP vWF

R. udzungwensis FMNH194127 KT348506 KT348370 KT348457 Udzungwa Mountains, Ndundulu Forest, Tanzania; �7.8045, 36.5059R. udzungwensis MTSN6000 KT348505 KT348371 KT348458 Udzungwa Mountains, Ndundulu Forest, Tanzania; �7.8036, 36.5059R. udzungwensis BMNH2007.7 KT000011 KT000020 KF202173 Udzungwa Mountains, Ndundulu Forest, Tanzania; �7.8045, 36.5059

E.J. Carlen et al. /Molecular Phylogenetics and Evolution 113 (2017) 150–160 155

were manually phased and given unique names (e.g. allele 1, allele2). Because so few SNPs were present in nuclear loci and allheterozygotes were restricted to a single change, phasing was donemanually. Geneious created alignments using the MAFFT v7.017alignment plugin (Katoh et al., 2002) for all of our assembledsequences, Rhynchocyon sequences downloaded from GenBank,and outgroup sequences. Alignments were checked by eye andexported for analysis. Duplicate haplotypes or allele sequencesfrom multiple individuals were identified and eliminated usingFaBox DNAcollapser v1.41 (Villesen, 2007). Aligned matrices foreach locus are available as Nexus files and published alongside thismanuscript as supplementary materials.

R. ch

R. chR. ch

0.025

*0.92

*85

**

Fig. 3. MrBayes phylogram for Rhynchocyon 12s16s mitochondrial region. Bayesian prepresented by an asterisk (*) above the branch. Maximum likelihood bootstraps equal tthe branch.

Each of the three independently segregating loci (12s16s, IRBP,vWF) were analyzed independently and as concatenated datasets.Because we were trying to assess the relationship of close relativesand the potential for gene flow, we analyzed each locus separatelyto specifically look for evidence of introgression and conflicting sig-nal, which may be ignored in analyses of concatenated matrices.

Phylogenetic analyses were performed using both maximumlikelihood and Bayesian approaches. First, Nexus files wereimported into PAUP⁄ v4.0b10 (Swofford, 2003) and sequenceswere partitioned into transfer RNAs and ribosomal RNAs for themitochondrial region, and into codon positions for the nuclear loci.We used MrModelTest v2.3 (Nylander, 2004) and the Akaike Infor-

R. cirnei cirnei CASMAM29344

E. edwardii

P. tetradactulysM. micus

rysopygus CASMAM24526

rysopygus CASMAM24525rysopugus FMNH153106

R. cirnei stuhlmanni M300R. cirnei reichardi MCZ43732

R. cirnei reichardi FMNH178010R. cirnei reichardi FMNH177823

R. cirnei reichardi FMNH171617R. cirnei reichardi FMNH171474

R. cirnei macrurus Douady et al. (2003)R. cirnei macrurus FMNH88204

R. cirnei shirensis AMNH161777R. cirnei cirnei CASMAM29351R. cirnei cirnei CASMAM29353

R. cirnei cirnei CASMAM29357R. cirnei cirnei CASMAM29355

R. cirnei cirnei CASMAM29358R. cirnei cirnei CASMAM29352R. cirnei cirnei CASMAM29345

R. udzungwensis FMNH194127R. udzungwensis BMNH2007.7

R. udzungwensis CASMAM28043

R. udzungwensis CASMAM28318R. udzungwensis MTSN6000

R. petersi petersi FMNH161395

R. petersi petersi FMNH192684

R. petersi petersi FMNH151213R. petersi petersi FMNH151214

R. petersi petersi TZ22767R. petersi adersi MCZ22829

R. petersi subsp. CASMAM28767R. petersi subsp. CASMAM29516

R. petersi petersi TA1835R. petersi petersi TZ22769

R. petersi petersi TA1818R. petersi petersi TZ22811R. petersi petersi TZ22779

R. petersi petersi TA1833

R. petersi petersi TZ22770R. petersi petersi TZ22778

R. petersi petersi TZ22775R. petersi petersi TZ22766

R. petersi petersi TZ22783

R. petersi petersi FMNH161311R. petersi petersi FMNH161312

R. petersi petersi TZ22774R. petersi petersi TZ22776

**

*65

**

**

***

81

700.76

0.6671

***81

*80

*74

**

**

*78

**

*44

*88 **

0.7744 *87

*63

0.8139

*74

osterior probabilities equal to or above 0.95 were considered significant and areo or above 90 were considered significant and are represented by an asterisk below

Table2

Distanc

ematrixfor12

s16s

mitoc

hond

rial

sequ

encessh

owingun

correctedpe

rcen

tdifferen

ces(p-distanc

e).T

hemea

nge

neticdistan

cean

dstan

dard

deviationis

show

nin

theup

perrigh

ttriang

le,a

ndrang

eis

show

nin

thelower

left

triang

le.

E. edwardii

M.m

icus

P. tetrad

actylus

R.

chrysopy

gus

R.c

irne

isthu

lman

niR.c

irne

ireicha

rdi

R.c

irne

imacrurus

R.cirne

ishrien

sis

R.cirne

icirnei

R.

udzu

ngwen

sis

R.p

etersi

adersi

R.p

etersisp

.R.p

etersi

E.ed

wardii

18.10

18.50

19.23±0.37

21.10

21.24±1.03

20.45±1.77

24.50

21.80±0.00

21.50±0.14

19.30

21.40±0.00

21.41±0.21

M.m

icus

18.10

17.70

20.77±0.06

21.50

21.64±0.25

21.30±0.57

25.60

21.90±0.05

22.18±0.27

20.80

21.70±0.00

21.75±0.18

P.tetrad

actylus

18.50

17.70

22.53±0.12

22.10

22.78±0.16

22.60±0.14

27.50

22.94±0.05

23.08±0.11

22.80

22.60±0.00

22.68±0.19

R.chrysop

ygus

19.50–

18.80

20.80–

20.70

22.60–

22.40

3.13

±0.15

3.01

±0.08

2.85

±0.10

8.35

±0.07

2.78

±0.07

3.00

±0.09

2.75

±0.07

2.95

±0.06

2.72

±0.10

R.cirne

isthu

lman

ni21

.10

21.50

22.10

3.00

–3.10

2.82

±0.18

2.50

±0.42

7.80

3.11

±0.04

3.30

±0.12

33.4±0.00

3.24

±0.07

R.cirne

ireicha

rdi

21.70–

19.40

21.80–

21.20

22.90–

22.50

3.00

–2.90

2.90

–2.50

1.51

±0.38

6.76

±0.13

1.58

±0.07

3.50

±0.43

2.54

±0.09

3.46

±0.30

3.37

±0.36

R.cirne

imacrurus

21.70–

19.20

21.70–

20.90

22.70–

22.50

2.90

–2.70

2.80

–2.20

2.00

–1.20

6.05

±0.07

1.04

±0.47

3.19

±0.78

2.20

±0.14

3.10

±0.81

2.88

±0.72

R.cirne

ishriensis

24.50

25.60

27.50

8.30

–8.40

7.8

7.00

–6.70

6.10

–6.00

6.09

±0/08

8.6±1.57

7.8

7.90

±0.00

7.74

±0.07

R.cirne

icirnei

21.8

22.00–

21.80

23.00–

22.90

2.90

–2.70

3.20

–3.10

1.70

–1.40

1.50

–0.50

6.20

–6.00

3.69

±0.15

2.29

±0.08

3.53

±0.07

3.37

±0.08

R.u

dzun

gwen

sis

21.70–

21.40

22.60–

22.00

23.20–

23.00

3.10

–2.90

3.50

–3.20

4.00

–2.60

4.20

–2.40

11.40–

7.80

4.00

–3.60

1.04

±0.05

1.24

±0.05

0.98

±0.08

R.p

etersi

adersi

19.30

20.80

22.80

2.80

–2.70

3.00

2.70

–2.50

2.30

–2.10

7.80

2.40

–2.20

1.10

–1.00

0.10

±0.00

0.22

±0.11

R.p

etersi

sp.

21.4

21.7

22.6

3.00

–2.90

3.4

3.60

–2.90

3.80

–2.40

7.9

3.70

–3.50

1.30

–1.20

0.10

0.42

±0.07

R.p

etersi

petersi

21.60–

21.00

21.90–

21.40

22.80–

22.30

3.10

–2.50

3.30

–3.10

3.60

–2.50

3.70

–2.10

7.90

–7.70

3.60

–3.30

1.10

–0.80

0.50

–0.10

0.50

–0.30

156 E.J. Carlen et al. /Molecular Phylogenetics and Evolution 113 (2017) 150–160

mation Criterion (Akaike, 1974) to assess the rate-specific model ofevolution for each partition (Supplementary Materials, Table 3).We performed Bayesian analysis using MrBayes v3.1.2 (Ronquistand Huelsenbeck, 2003). Bayesian analysis was run for 10 millionMarkov Chain Monte Carlo (MCMC) generations, sampling treesand parameters every 1000th generation. The first 25% of the gen-erations sampled were discarded as burnin. We performedmaximum-likelihood analysis using Random Axelerated MaximumLikelihood (RAxML) v7.2.6 using the GTR + C model (Stamatakis,2006). The ‘‘autoMRE” command, which is a bootstrap convergencetest, was used to determine when a sufficient number of bootstrapreplicates had been reached (Pattengale et al., 2010).

To test the robustness of the results and the impact of missingdata, we removed any individuals with over 50% missing data fromthe matrix, and repeated the analysis on the reduced matrix. Allanalyses were performed on the Phylocluster computer at CAS.Support for each node was estimated using Bayesian posteriorprobabilities in MrBayes (Ronquist and Huelsenbeck, 2003) andbootstrap analysis in RAxML (Stamatakis, 2006). We created agenetic distance matrix for the 12s16s region in Geneious v7.1.4(Kearse et al., 2012) by subtracting the percent identity providedin the multiple alignment from 100 and averaging across individu-als for the same taxon pairs. Nuclear loci were visualized asunrooted TCS allele networks (Clement et al., 2000) using PopARTv1 (Leigh et al., 2013).

To assess the support for species delimitation, we used themultispecies coalescent model in the Bayesian reverse-jumpMCMC program BPP, version 3.3 (Yang, 2015). We assessed thephylogenetic support for the four recognized Rhynchocyon species(R. chrysopygus, R. cirnei, R. petersi, and R. udzungwensis), but weadditionally wanted to assess support for R. c. stuhlmanni as adistinct species. To do this, we utilized the A11 model to performjoint species delimitation and species tree estimation, as unguideddelimitation is preferred over delimitation using a fixed guidetree and can sometimes show strong species support even whenmulti-locus phylogenies are uncertain (Yang, 2015). Control filescontaining the important parameter settings and Bayesian priorsare provided in Supplementary Material file: BPP Analyses.

3. Results

Final aligned sequence length for 12s16s equaled 2685 nucleo-tide base pairs representing 48 specimens across 10 giant sengiand 3 outgroup taxa. Final aligned sequence length for IRBPequaled 976 base pairs representing 45 specimens across 8 giantsengi taxa. Final aligned sequence length for vWF equaled 1043base pairs representing 45 specimens across 8 giant sengi taxa.Specimens with missing data were mostly from degraded museumskin samples from which our genes were difficult to amplify orsequence (see Table 1). Bayesian and maximum likelihood analysesof the mitochondrial locus 12s16s (Fig. 3) recovered similar treeswith consistent support for nodes. We considered node supportas significant if the Bayesian posterior probability (Bpp) was equalto or greater than 0.95 and the maximum likelihood bootstrap(mlb) support was equal to or greater than 90. In our 12s16s tree(Fig. 3), there was good phylogenetic resolution at the species leveland even at the subspecies level for almost every taxon.

The average pairwise genetic distance matrix for the 12s16sregion shows percent differences between 0.1% and 8.8%(�x = 2.4 ± 1.8) among Rhynchocyon taxa (Table 2). The greatest dif-ferences are between R. c. shirensis and other taxa. However, we areskeptical of these values because R. c. shirensis has approximately1400 fewer bases than other taxa due to poorly preserved DNA,and missing data may contribute to unusual estimates. Excluding

Fig. 4. TCS allelic networks for Rhynchocyon nuclear loci IRBP and vWF. The size of the circle is proportional to the number of alleles sampled of that haplotype. Black circlesrepresent inferred alleles that were not recovered in this analysis and ‘n’ represents the number of alleles recovered for each haplotype.

E.J. Carlen et al. /Molecular Phylogenetics and Evolution 113 (2017) 150–160 157

R. c. shirensis, the remaining Rhynchocyon taxa have average pair-wise genetic distances ranging from 0.2% to 3.7% (�x = 2.0 ± 1.4).

Nuclear loci IRBP and vWF were chosen based on previous workwith Macroscelididae (Douady et al., 2003; Dumbacher et al., 2014;Lawson et al., 2013; Smit et al., 2011; Springer et al., 1997). How-ever, these loci exhibited low variation among the samples ana-lyzed and we recovered only seven single nucleotidepolymorphisms (SNPs) across the 976 IRBP bases and eleven SNPsacross the 1043 vWF bases. Thus, we chose to visualize nuclear lociIRBP and vWF as individual allelic networks (Fig. 4).

Within the IRBP allele network, R. petersi alleles cluster togetherwith all but one allele from R. cirnei subspecies (n = 74 sampledalleles). Most individuals of these taxa are represented by a singleshared allele (n = 70 sampled alleles) and only three other allelesthat are one nucleotide substitution different (n = 4). R. c. stuhl-manni, however, is separated from the remaining R. cirnei sub-species by three nucleotide changes, and is the most distant IRBPallele from other R. cirnei found in the genus. Additionally, R.chrysopygus and R. udzungwensis share an allele which differs byone nucleotide change from the R. c. stuhlmanni allele and theunique R. chrysopygus allele.

The vWF locus contains more genetic variation and taxonomicstructure. In the vWF network (Fig. 4), only one allele is presentin multiple taxa. This allele is the most common allele overall(n = 57 alleles), and it is shared by R. c. cirnei, R. p. petersi, andthe two captive R. petersi from the Houston Zoo (CAS MAM28767 and CAS MAM 29516). Three additional alleles from R. c.macrurus and R. petersi diverged from this most common alleleby only one nucleotide change. The four remaining taxa, (R. c. reich-ardi, R. c. stuhlmanni, R. chrysopygus, and R. udzungwensis), eachappear relatively distinct; each taxon has unique alleles that are

at least two nucleotide changes to any allele belonging to anothertaxon. Allele networks for both nuclear loci clearly show the dis-tinctness of R. c. stuhlmanni, R. chrysopygus, and R. udzungwensis.

The results of the mitochondrial and nuclear analyses weremostly compatible. Discordance among the analyses came fromthe close affinity of R. cirnei and R. petersi in the nuclear genomeand shared alleles in both IRBP and vWF (Fig. 4), and yet thesetwo taxa were distant clades in the mitochondrial analysis of12s16s (Fig. 3). In the mitochondrial analysis, R. p. petersiwas moreclosely related to R. udzungwensis than any other taxon. All threeloci supported the phylogenetic distinctness of R. c. stuhlmanni.

Our analysis showed that the Boni tissue sample was 100%identical to R. chrysopygus for the mitochondrial locus and vWF.For IRBP, the Boni individual was heterozygous with one allelematching a R. chrysopygus allele, while the other allele was newto our analysis and one nucleotide change different from the alleleshared by R. udzungwensis and R. chrysopygus. Thus, the tissue wesequenced and analyzed from the Boni population is indistinguish-able from R. chrysopygus. We have not included these data in thefigures.

BPP analyses using the entire dataset suffered from mixingproblems such that initial conditions affected the outcome. Specif-ically, searches starting at or near the one species model (0000)tended to get stuck in that model, whereas models that initiallyincluded multiple species progressed toward strong support forthe five species model (1111). Although the manual suggestedsome potential causes (e.g. too many loci, inappropriate priors),we explored these options without improving the mixing. Weadditionally explored the role of missing data, and by eliminatingindividuals with any locus completely missing, we found thatmixing and rjMCMC behavior improved. This dataset included 44

158 E.J. Carlen et al. /Molecular Phylogenetics and Evolution 113 (2017) 150–160

individuals (R. chrysopygus, n = 1; R. cirnei, n = 15, R. c. stuhlmanni,n = 1; R. petersi, n = 23; and R. udzungwensis, n = 4), and was runfive times to ensure consistent outcomes. The results consistentlysupported delimiting all five species (including elevating R. c. stuh-manni to full species) with posterior probabilities >0.988 per run(n = 5 runs).

4. Discussion

With nearly complete taxon sampling (we lack R. c. hendersoni),our analysis confirms that earlier taxonomists (Corbet and Hanks,1968; Rovero et al., 2008) accurately inferred taxonomic groupswithin Rhynchocyon using pelage color patterns and geographicrange. Below we discuss our findings in relation to previous workon Rhynchocyon and include a revised taxonomy.

4.1. Clarifying the taxonomic status of ambiguous GenBank sequences

Douady et al. (2003) posted three sequences for Rhynchocyon sp.on GenBank: 12s16s (GenBank accession number AY310880), IRBP(AY310894), and vWF (AY310887). All three sequences clusterwith R. cirnei in our analyses, in contrast to Smit et al. (2011) thatsuggested these are R. chrysopygus. Thus, we wanted to check forother data confirming the taxonomic identity of this specimen.GenBank entries share a single extraction number (CJD-2003). Inhis dissertation, Douady (2001) lists two tissues as the source ofgenetic data for his Rhynchocyon sp., (tissue numbers T-1853, T-1854), from the personal collection of François Catzeflis at theUniversité Montpellier, France, and provides the collection localityfor both tissues as Chingulungulu (Tanzania). Douady (2001) doesnot report which tissue was sequenced and posted on GenBank,but for our purposes it makes little difference because both vou-cher specimens (collected by Herwig Leirs and Walter Verheyen,Royal Museum of Central Africa, Tervuren, Belgium, catalog num-bers 96.037-M-5388 and 96.037- M-5390) came from the samelocality (10�440S, 38�330E), and a visual inspection of images ofthe two specimens indicates they were the same taxon (F. Catezflisand H. Leirs, pers. comm.). Based on pelage color and pattern, dis-tribution, and our sequence data, we conclude that the GenBanksequence is from R. c. macrurus and not R. chrysopygus as proposedby Smit et al. (2011).

4.2. Origins of captive populations

It is not known where the founders of the captive population ofR. petersi at zoos in the United States (Baker et al., 2005) came from(unpublished correspondence, K. Lengel, P. Riger, and S. Eller).Based on pelage coloration, the sengis are obviously R. petersi,but the two subspecies have different geographic distributions,with R. p. petersi from mainland Tanzania and southeastern Kenya,and R. p. adersi from the islands of Mafia and Zanzibar off the coastof Tanzania. For captive breeding purposes, and if reintroductionsshould be contemplated in the future, it would be important toknow the provenance of the captive population to maintain thegenetic integrity of both wild and captive populations. We ana-lyzed the DNA of two captive individuals from the Houston Zoo(Table 1), and these clustered with R. p. adersi in the 12s16s mito-chondrial phylogram (Fig. 3). We were unable to confirm the clus-tering at the nuclear loci because we were unable to sequencenuclear DNA from any confirmed R. p. adersi. However, our analysissuggests that the zoo specimens were originally taken from R. p.adersi exported from Zanzibar or Mafia islands. Because we hadonly a single R. p. adersi sample from the wild population, andbecause it genetically clustered well within the available variationof the R. p. petersi clade, we regard these results as preliminary.

4.3. Hybridization

Lawson et al. (2013) presented data consistent with ancientintrogression between R. c. reichardi and R. udzungwensis, wherethe distribution of the two taxa meet in the Udzungwa Mountainsof Tanzania, calling into question the genetic boundaries of thesetwo taxa. We found no evidence of introgression between anyRhynchocyon species, but none of our samples were from adjoiningpopulations like those of Lawson et al. (2013). The differencesbetween our two studies are likely explained by the differencesin geographical sampling and perhaps the depth of sampling.Lawson et al. (2013) sampled extensively across a narrow range,targeting the contact zone of R. c. reichardi and R. udzungwensis.We sampled shallowly across a broad range, mostly away fromcontact zones. Therefore, if introgression occurs at contact zones,we were less likely to detect it. Indeed, our data suggest that wide-spread gene flow and panmixia does not occur in Rhynchocyon.

Furthermore, our 12s16s phylogeny shows that R. c. reichardiand R. udzungwensis are not sister taxa, with R. udzungwensis beingmore closely related to R. petersi than the R. cirnei clade (Fig. 3).Observations of hybridization between non-sister species has beendocumented in other groups (Dasmahapatra et al., 2007; Goodet al., 2003; Larsen et al., 2010; McKitrick and Zink, 1988). Hybridssometime occur when non-sister species have overlapping rangesor historical contact zones, such as the contact zone between R.udzungwensis and R. c. reichardi. Thus, studies looking for evidenceof Rhynchocyon introgression should sample in areas where histor-ical contact between species may have occurred, although thesemay be exceedingly difficult to find given the disappearance of for-est habitats in Africa.

4.4. Current taxonomic status of Rhynchocyon taxa

In their revision of Macroscelididae, Corbet and Hanks (1968)described a new subspecies of R. cirnei, R. c. shirensis, from the ShireValley of Malawi, based on a distinct pelage pattern. However, Coalsand Rathbun (2013) examined additional museum specimens andobserved that the pelage of the Malawi subspecies appeared to bewithin the variation seen inR. c. cirnei specimens fromMozambique.We have only one sample of R. c. shirensis in ourmitochondrial anal-ysis and no samples in our nuclear analyses; nonetheless, in the12s16s phylogeny, R. c. shirensis falls within the R. cirnei clade(Fig. 3) and does not cause any taxa to be paraphyletic, therefore,we recommend continuing to treat R. c. shirensis as a subspecies ofR. cirnei pending additional sampling and analyses.

In Andanje et al. (2010) suggested a potentially new species ofRhynchocyon from the Dodori and Boni national reserves on thenorthern coast of Kenya. A voucher specimen was collected andplaced at the National Museums of Kenya (NMK169427), and tis-sue from this voucher was sent to CAS by the Kenya Wildlife Ser-vice. The sequences that we obtained from the tissue wereidentical to R. chrysopygus at 12s16s, vWF, and one of two allelesat IRBP. These data suggest that the specimen we sequenced isgenetically very similar to, or perhaps a form of, R. chrysopygus.This is surprising given the very different pelage color and patternsbetween these two allopatric forms (Andanje et al., 2010). More-over, our work suggests that dorsal pelage pattern and colorationare useful taxonomic characters for other Rhynchocyon taxa.Because our results are based upon a single tissue specimen, weare reluctant to draw any conclusions regarding this specimenand the sequences without examining the voucher. More datashould be collected and analyzed before any conclusions can bemade about the taxonomic status of this morphologically uniquegiant sengi.

Along with the six R. cirnei subspecies, Corbet and Hanks (1968)suggested a potential seventh subspecies based on a single speci-

E.J. Carlen et al. /Molecular Phylogenetics and Evolution 113 (2017) 150–160 159

men collected in northeastern Mozambique, but noted that thisspecimen might be an intermediate between R. c. cirnei and R. c.macrurus based on tail coloration. To investigate this potential sub-species, Coals and Rathbun (2013) collected eight Rhynchocyonspecimens from northeastern Mozambique and compared thepelage of their specimens with several R. c. cirnei individuals,including two topotypes. They concluded that the variation inpelage color and pattern within the distribution of R. c. cirnei doesnot justify designation of any new subspecies, and additionallyquestioned the validity of R. c. shirensis pending genetic analyses.Although our genetic analysis of R. c. shirensis suggests it is distinctfrom R. c. cirnei, all of our R. c. cirnei samples are from northernMozambique, and we do not have tissue from R. c. cirnei topotypesfrom southernMozambique. Because we had only one R. c. shirensisspecimen in our analysis, and no R. c. cirnei from nearby southernMozambique where the type specimens originated, we are unableto assess Corbet and Hanks’s (1968) question as to whether thenorthern Mozambique form of R. c. cirnei might be genetically dis-tinct from the southern form from southern Mozambique andMalawi.

Our molecular data suggest that R. c. stuhlmanni could bereturned to full species, as provisionally proposed by Corbet andHanks (1968) based upon short nasals and disjunct range. The12s16s phylogeny (Fig. 3) shows strong support for R. c. stuhlmannias a distinct lineage that is sister to all other R. cirnei subspecies.The mean distance matrix for 12s16s (Table 2) shows R. c. stuhl-manni as at least 2% divergent from other R. cirnei, while theremaining R. cirnei subspecies show among subspecies divergencesbetween 1% and 1.6%. Moreover, the nuclear allele networks(Fig. 3) show additional support for the uniqueness of R. c. stuhl-manni and support elevating it to full species. In both the IRBPand vWF allele networks, R. c. stuhlmanni has a unique allele thatis not shared by any other taxa. Furthermore, the R. c. stuhlmanniallele in the IRBP network is three steps away from the other R. cir-nei subspecies, and closer to an allele shared by R. chrysopygus andR. udzungwensis. Results of BPP analyses additionally corroboratedelimiting R. c. stuhlmanni as a distinct taxon. Thus R. c. stuhlmanniis morphologically, geographically, and genetically distinct fromother R. cirnei. However, the cline described by Corbet and Hanks(1968) needs genetic examination, as does the cline they describein southeastern Tanzania for R. c. macrurus.

The phylogenetic data and taxonomic revision that we presenthere will facilitate a future detailed treatment of Rhynchocyon phy-logeography. For example, it has been proposed that large riversand their flood plains, as well as lowland ground-water forests,are important limiting factors in the historical and current distri-bution of Rhynchocyon (Corbet and Hanks, 1968; Andanje et al.,2010; Coals and Rathbun, 2013; Rathbun, 2009). It is also possiblethat the Rift Valley lakes and highlands were prehistorically impor-tant vicariant factors, although the current distribution of Rhyn-chocyon taxa suggests that neither the lakes nor elevationcompletely account for current distributions (Fig. 2, www.sengis.org/distribution). An analysis of the phylogeography of Rhyn-chocyon will need to include a careful assessment of the diversifi-cation of other faunal groups in Africa and the likelyenvironmental factors involved, such as climate shifts, tectonics,forest fragmentation, river meanderings, and hydrological basinshifts (Kingdon, 1989; Stanley et al., 2005; deMenocal, 2004;Lawson, 2010; Dimitrov et al., 2012; Taylor et al., 2009; Fjeldsaand Bowie, 2008).

5. Conclusions

Based on our genetic analysis we recommend the following tax-onomic treatment for giant sengis:

Family: Macroscelididae Bonaparte, 1838Subfamily: RhynchocyoninaeGenus: Rhynchocyon Peters, 1847Rhynchocyon cirnei Peters, 1847Rhynchocyon cirnei cirnei Peters, 1847Rhynchocyon cirnei shirensis Corbet and Hanks, 1968Rhynchocyon cirnei reichardi Reichenow, 1886Rhynchocyon cirnei hendersoni Thomas, 1902Rhynchocyon cirnei macrurus Günther, 1881

Rhynchocyon stuhlmanni Matschie, 1893Rhynchocyon petersi Bocage, 1880Rhynchocyon petersi petersi Bocage, 1880Rhynchocyon petersi adersi Dollman, 1912

Rhynchocyon chrysopygus Günther, 1881Rhynchocyon udzungwensis Rathbun & Rovero, 2008

Several important taxonomic issues remain to be resolved that willrequire further research using molecular genetics in conjunctionwith morphology and distribution data. These issues include 1.whether the Rhynchocyon from the northern coastal area of Kenyarepresents a new species, 2. whether R. c. hendersoni is a valid sub-species rather than only a relatively minor geographic (high eleva-tion) variant of R. c. reichardi, 3. whether R. c. reichardi should bereturned to full species status, 4. whether R. c. shirensis representsa minor variant within R. cirnei and thus should not be a subspecies,and 5. the genetic nature of the geographic variation in pelage pat-tern in southeastern Tanzania (R. c. macrurus), the Congo Basin (R.stuhlmanni), and Mozambique and southern Malawi (R. c. cirneiand shirensis). In any case, the prediction that continued revisionsof the taxonomy of Corbet and Hanks (1968) would result in greaterrecognized sengi diversity (Rathbun, 2009) is being born out withthe recent revision of the genus Macroscelides to include three spe-cies (Dumbacher et al., 2014), and the creation of a new genus forthe North African sengi, Petrosaltator rozeti (Dumbacher et al.,2016), and in this paper with the resurrection of R. stuhlmanni.

Acknowledgements

Our research was financially supported by a National ScienceFoundation grant to LEO (DEB-1120904). Additionally, we thankThe California Academy of Sciences, the Biology Department andGraduate Student Council in Biology at San Francisco State Univer-sity, the Society for the Study of Evolution, and the Society for Inte-grative and Comparative Biology for financial support. We thank B.R. Agwanda of the National Museums of Kenya who captured andprepared the Boni Rhynchocyon, S. Andanje of the Kenya WildlifeService who personally imported Rhynchocyon tissue into the Uni-ted States of America from the wildlife service’s collection, and S.Musila of the National Museums of Kenya who encouraged us toinclude the Boni Rhynchocyon in our analyses. Thanks to K. Conso-late, who provided us with tissues of R. c. stuhlmanni. We are grate-ful for specimens and photographs provided by N. Duncan at theAmerican Museum of Natural History, and specimens providedby J. Chupasko at the Harvard Museum of Comparative Zoology.We thank K. Hildebrandt, O. Carmi, and A. Sellas who assisted withlaboratory work. Thanks to H. A. Smit-Robinson who shared adviceand details of her previously published work on Rhynchocyon. F.Catzeflis, L. Herwig, andW.Wendelen helped determine the prove-nance of the Douady et al. (2003) specimen, including providingphotographs of specimens in the collection of the Royal Museumof Central Africa. Thanks to K. Lengel and S. Eller, at the Philadel-phia Zoo, and P. Riger, at the Houston Zoo who provided informa-tion about Rhynchocyon zoo specimens. Lastly, we dedicate thispaper to our coauthor, friend, and colleague W. T. Stanley, who

160 E.J. Carlen et al. /Molecular Phylogenetics and Evolution 113 (2017) 150–160

died suddenly and unexpectedly in November 2015 while conduct-ing research in Ethiopia.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.ympev.2017.05.012.

References

Akaike, H., 1974. A new look at the statistical model identification. IEEE Trans.Autom. Control 19, 716–723. http://dx.doi.org/10.1109/TAC.1974.1100705.

Andanje, S., Agwanda, B.R., Ngaruiya, G.W., Amin, R., Rathbun, G.B., 2010. Sengi(elephant-shrew) observations from northern coastal Kenya. J. East African Nat.Hist. 99, 1–8. http://dx.doi.org/10.2982/028.099.0101.

Baker, A.J., McCafferty, K., Hellmuth, H., 2005. Black-and-rufous sengi (Rhynchocyonpetersi) at the Philadelphia Zoo. Afrotherian Conservation - Newsletter of theIUCN-SCC Afrotheria Specialist Group 3, 6–7.

Clement, M., Posada, D., Crandall, K.A., 2000. TCS: a computer program to estimategene genealogies. Mol. Ecol. 9, 1657–1659. http://dx.doi.org/10.1046/j.1365-294X.2000.01020.x.

Coals, P.G.R., Rathbun, G.B., 2013. The taxonomic status of giant sengis (genusRhynchocyon) in Mozambique. J. East African Nat. Hist. 101, 241–250. http://dx.doi.org/10.2982/028.101.0203.

Corbet, G.B., Hanks, J., 1968. A revision of the elephant-shrews, familyMacroscelididae. Bull. Br. Museum (Natural Hist.) Zool. 16, 45–111.

Dasmahapatra, K.K., Silva-Vásquez, A., Chung, J., Mallet, J., 2007. Genetic analysis ofa wild-caught hybrid between non-sister Heliconius butterfly species. Biol. Lett.3, 660–663. http://dx.doi.org/10.1098/rsbl.2007.0401.

Davies, T.J., Buckley, L.B., 2011. Phylogenetic diversity as a window into theevolutionary and biogeographic histories of present-day richness gradients formammals. Philos. Trans. R. Soc. B 366, 2414–2425. http://dx.doi.org/10.1098/rstb.2011.0058.

deMenocal, P.B., 2004. African climate change and faunal evolution during thePliocene-Pleistocene. Earth Planet. Sci. Lett. 220, 3–24.

Dimitrov, D., Nogués-Bravo, D., Scharff, N., 2012. Why do tropical mountainssupport exceptionally high biodiversity? The eastern arc mountains and thedrivers of saintpaulia diversity. PLoS ONE 7 (11), e48908. http://dx.doi.org/10.1371/journal.pone.0048908.

Douady, C., 2001. Molecular Phylogenetics of the Insectivora PhD diss. The Queen’sUniversity of Belfast.

Douady, C.J., Catzeflis, F., Raman, J., Springer, M.S., Stanhope, M.J., 2003. The Saharaas a vicariant agent, and the role of Miocene climatic events, in thediversification of the mammalian order Macroscelidea (elephant shrews).Proc. Natl. Acad. Sci. U.S.A. 100, 8325–8330. http://dx.doi.org/10.1073/pnas.0832467100.

Dumbacher, J.P., Rathbun, G.B., Smit, H.A., Eiseb, S.J., 2012. Phylogeny and taxonomyof the round-eared sengis or elephant-shrews, genus Macroscelides (Mammalia,Afrotheria, Macroscelidea). PLoS ONE 7, e32410. http://dx.doi.org/10.1371/journal.pone.0032410.

Dumbacher, J.P., Rathbun, G.B., Osborne, T.O., Griffin, M., Eiseb, S.J., 2014. A newspecies of round-eared sengi (genus Macroscelides) from Namibia. J. Mammal.95, 443–454. http://dx.doi.org/10.1644/13-MAMM-A-159.

Dumbacher, J.P., Carlen, E.J., Rathbun, G.B., 2016. Petrosaltator gen. nov., a newgenus replacement for the North African sengi Elephantulus rozeti(Macroscelidea; Macroscelididae). Zootaxa 4136, 567–579.

Faith, D.P., 1992. Conservation evaluation and phylogenetic diversity. Biol. Conserv.61, 1–10. http://dx.doi.org/10.1016/0006-3207(92)91201-3.

Fjeldsa, J., Bowie, R.C.K., 2008. New perspectives on the origin and diversification ofAfrica’s forest avifauna. Afr. J. Ecol. 46, 235–247.

FitzGibbon, C.D., 1997. The adaptive significance of monogamy in the golden-rumped elephant-shrew. J. Zool. 242, 167–177. http://dx.doi.org/10.1111/j.1469-7998.1997.tb02937.x.

Good, J.M., Demboski, J.R., Nagorsen, D.W., Sullivan, J., 2003. Phylogeography andintrogressive hybridization: chipmunks (genus Tamias) in the northern RockyMountains. Evolution 57, 1900–1916. http://dx.doi.org/10.1111/j.0014-3820.2003.tb00597.x.

Katoh, K., Misawa, K., Kuma, K., Miyata, T., 2002. MAFFT: a novel method for rapidmultiple sequence alignment based on fast Fourier transform. Nucl. Acids Res.30, 3059–3066. http://dx.doi.org/10.1093/nar/gkf436.

Kearse, M., Moir, R., Wilson, A., Stones-Havas, S., Cheung, M., Sturrock, S., Buxton, S.,Cooper, A., Markowitz, S., Duran, C., Thierer, T., Ashton, B., Meintjes, P.,Drummond, A., 2012. Geneious Basic: an integrated and extendable desktopsoftware platform for the organization and analysis of sequence data.Bioinformatics 28, 1647–1649. http://dx.doi.org/10.1093/bioinformatics/bts199.

Kingdon, J., 1989. Island Africa – The Evolution of Africa’s Rare Animals and Plants.Princeton University Press, New Jersey. 287 pp.

Larsen, P.A., Marchán-Rivadeneira, M.R., Baker, R.J., 2010. Natural hybridizationgenerates mammalian lineage with species characteristics. Proc. Natl. Acad. Sci.U.S.A. 107, 11447–11452. http://dx.doi.org/10.1073/pnas.1000133107.

Lawson, L.P., 2010. The discordance of diversification: evolution in the tropical-montane frogs of the Eastern Arc Mountains of Tanzania. Mol. Ecol. 19, 4046–4060.

Lawson, L.P., Vernesi, C., Ricci, S., Rovero, F., 2013. Evolutionary history of the grey-faced sengi, Rhynchocyon udzungwensis, from Tanzania: a molecular and speciesdistribution modeling approach. PLoS ONE 8, e72506. http://dx.doi.org/10.1371/journal.pone.0072506.

Leigh, J., Bryant, D., Steel, M., 2013. PopART (Population Analysis with ReticulateTrees). <http://www.popart.otago.ac.nz/downloads.shtml> (Accessed 22February 2015).

McKitrick, M.C., Zink, R.M., 1988. Species concepts in ornithology. Condor 90, 1–14.Novacek, M., 1984. Evolutionary stasis in the elephant-shrew, Rhynchocyon. In:

Eldredge, N., Stanley, S.M. (Eds.), Living Fossils. Springer Verlag, New York, NewYork, pp. 4–22.

Nylander, J.A.A., 2004. MrModeltest v2. Program Distributed by the Author.Evolutionary Biology Centre, Uppsala University.

Pattengale, N., Alipour, M., Bininda-Emonds, O., Moret, B., Stamatakis, A., 2010. Howmany bootstrap replicates are necessary? J. Comput. Biol. 17, 337–354. http://dx.doi.org/10.1089/cmb.2009.017.

Puttick, M.N., Thomas, G.H., 2015. Fossils and living taxa agree on patterns of bodymass evolution: a case study with Afrotheria. Proc. R. Soc. B-Biol. Sci. 282,20152023. http://dx.doi.org/10.1098/rspb.2015.2023.

Rathbun, G.B., 1979. The social structure and ecology of elephant-shrews. Z.Tierpsychol. (Supplement 20), 1–77.

Rathbun, G.B., 2009. Why is there discordant diversity in sengi (Mammalia:Afrotheria: Macroscelidea) taxonomy and ecology? Afr. J. Ecol. 47, 1–13.

Rathbun, G.B., Kingdon, J., 2006. The etymology of ‘sengi.’ Afrotherian ConservationNewsletter IUCN/SSC Afrotheria Specialist Gr. 4, 14–15.

Ronquist, F., Huelsenbeck, J.P., 2003. MrBayes 3: Bayesian phylogenetic inferenceunder mixed models. Bioinformatics 19, 1572–1574. http://dx.doi.org/10.1093/bioinformatics/btg180.

Rovero, F., Rathbun, G.B., Perkin, A., Jones, T., Ribble, D.O., Leonard, C., Mwakisoma,R.R., Doggart, N., 2008. A new species of giant sengi or elephant-shrew (genusRhynchocyon) highlights the exceptional biodiversity of the UdzungwaMountains of Tanzania. J. Zool. 274, 126–133. http://dx.doi.org/10.1111/j.1469-7998.2007.00363.x.

Sabuni, C.A., Van Houtte, N., Gryseels, S., Maganga, S.L.S., Makundi, R.H., Leirs, H.,Goüy de Bellocq, J., 2016. Genetic structure and diversity of the black and rufoussengi in Tanzanian coastal forests. J. Zool. 1–9. http://dx.doi.org/10.1111/jzo.12384.

Smit, H., 2008. Phylogeography of Three Southern African Endemic Elephant-Shrews and a Supermatrix Approach to the Macroscelidea PhD diss.Stellenbosch University.

Smit, H.A., Robinson, T.J., van Vuuren, B.J., 2007. Coalescence methods reveal theimpact of vicariance on the spatial genetic structure of Elephantulus edwardii(Afrotheria, Macroscelidea). Mol. Ecol. 16, 2680–2692. http://dx.doi.org/10.1111/j.1365-294X.2007.03334.x.

Smit, H.A., Robinson, T.J., Watson, J., van Vuuren, B.J., 2008. A new species ofelephant-shrew (Afrotheria: Macroscelidea: Elephantulus) from South Africa. J.Mammal. 89, 1257–1268.

Smit, H.A., van Vuuren, J., O’Brien, P.C.M., Ferguson-Smith, M., Yang, F., Robinson, T.J., 2011. Phylogenetic relationships of elephant-shrews (Afrotheria,Macroscelididae). J. Zool. 284, 133–143. http://dx.doi.org/10.1111/j.1469-7998.2011.00790.x.

Smith, H.F., Parker, W., Kotzé, S.H., Laurin, M., 2013. Multiple independentappearances of the cecal appendix in mammalian evolution and aninvestigation of related ecological and anatomical factors. C.R. Palevol 12,339–354. http://dx.doi.org/10.1016/j.crpv.2012.12.001.

Smith, H.F., Parker, W., Kotzé, S.H., Laurin, M., 2016. Morphological evolution of themammalian cecum and cecal appendix. C.R. Palevol 16, 39–57. http://dx.doi.org/10.1016/j.crpv.2016.06.001.

Springer, M.S., Cleven, G.C., Madsen, O., de Jong, W.W., Waddell, V.G., Amrine, H.M.,Stanhope, M.J., 1997. Endemic African mammals shake the phylogenetic tree.Nature 388, 61–64. http://dx.doi.org/10.1038/40386.

Stamatakis, A., 2006. RAxML-VI-HPC: maximum likelihood-based phylogeneticanalyses with thousands of taxa and mixed models. Bioinformatics 22, 2688–2690. http://dx.doi.org/10.1093/bioinformatics/btl446.

Stanley, W.T., Rogers, M.A., Hutterer, R., 2005. A new species of Congosorex from theEastern Arc Mountains, Tanzania, with significant biogeographical implications.J. Zool., London 265, 269–280.

Swofford, D.L., 2003. PAUP⁄. Phylogenetic Analysis Using Parsimony (⁄and OtherMethods). Version 4. Sinauer Associates, Sunderland, Massachusetts.

Taylor, P.J., Maree, S., Van Sandwyk, J., Kerbis Peterhans, J.C., Stanley, W.T.,Verheyen, E., Kaliba, P., Verheyen, W., Kaleme, P., Bennett, N.C., 2009.Speciation mirrors geomorphology and palaeoclimatic history in Africanlaminate-toothed rats (Muridae: Otomyini) of the Otomys denti and Otomyslacustris species-complexes in the ‘Montane Circle’ of East Africa. Biol. J. Lin. Soc.96, 913–941.

Villesen, P., 2007. FaBox: an online toolbox for FASTA sequences. Mol. Ecol. Notes 7,965–968. http://dx.doi.org/10.1111/j.1471-8286.2007.01821.

Yang, Z., 2015. The BPP program for species tree estimation and speciesdelimitation. Curr. Zool. 61, 854–865.